Investing.com -- Morgan Stanley raised its price targets on Nvidia and Broadcom after a series of meetings in Asia and the U.S. pointed to broad-based strength in AI demand and tighter supply conditions across the semiconductor ecosystem.

The bank sees both companies as well-positioned to ramp meaningfully next year as customers struggle to secure enough AI hardware to meet accelerating workloads.

Analysts led by Joseph Moore said they “continue to see Nvidia maintaining dominant market share,” arguing that perceived competitive threats appear “overstated” and that customers’ biggest concern over the next year is securing sufficient supply, particularly of the upcoming Vera Rubin platform.

Nvidia’s data center revenues remain supply-constrained through 2026, and Morgan Stanley now sees the company on track to approach the revenue levels outlined by management earlier this year. The bank raised its target price to $250 from $235.

“We still are below the "$500 bn in 5 quarters" voiced by the CEO at the GTC event, but clearly the situation is strong,” Moore said in a Monday note.

Moreover, the analyst said checks support stronger-than-expected demand for Google’s Tensor Processing Unit (TPU) supply chain, designed and sold by Broadcom. Multiple suppliers across analog, memory and ODMs pointed to upward revisions in TPU builds, with the biggest increases in 2027.

“We remain Overweight (OW) AVGO and we are encouraged to see this - and raising numbers - but we would introduce a few caveats,” Moore wrote.

He also noted that Google is pushing ahead with a homegrown TPU variant in partnership with Mediatek, describing it as a potential long-term risk but not one that meaningfully alters its base-case view.

Morgan Stanley hiked its Broadcom price target to $443 from $409, reflecting higher ASIC revenue expectations in 2026 and 2027.

The Wall Street giant now models $27.2 billion in ASIC revenue next year and $59.5 billion the following year, supported by stronger CoWoS volume projections of 3.2 million units in 2026 and 5 million in 2027.

The surge in AI demand is straining not only back-end capacity such as CoWoS and high-bandwidth memory but also front-end wafers, Moore pointed out. Tight supply across 3-, 4- and 5-nanometer nodes is reinforcing a bullish setup for the broader group, with the industry requiring more capacity additions.

Memory markets are also tightening sharply, with cloud buyers showing what the analysts describe as a “gold rush purchasing mentality.”

Across the ecosystem, demand for general-purpose server CPUs remains strong, with AMD continuing to gain share as Intel’s supply growth falls short.

Moore also flagged ongoing progress at Astera Labs and limited exposure to China localization efforts for the largest U.S. players.